Implicit conversion from float -65.6532663316583 to int loses precision